Strange routing on HP-UX 11.11

[Haim Lichaa]

I have 2 network interfaces on two RP4440's, one of which is connected to a closed segment.
When ever I remote copy a file (push) from one RP to the other, over the closed segment it works fine.
The problem is copying a file to (pull) my local RP from the remote RP (just hangs).

for example:
#rcp tmpfile remoteserver:/tmpfile (works)
Now I want to get the file back
#rcp remoteserver:/tmpfile /. (doesn't work, just hangs).

Here is my netconf file
#
INTERFACE_NAME[0]=lan1
#
IP_ADDRESS[0]=172.23.4.144
SUBNET_MASK[0]=255.255.255.0
BROADCAST_ADDRESS[0]=172.23.4.255
INTERFACE_STATE[0]=up
DHCP_ENABLE[0]=0
#
INTERFACE_NAME[1]=lan0
IP_ADDRESS[1]=172.23.182.21
SUBNET_MASK[1]=255.255.255.0
BROADCAST_ADDRESS[1]=172.23.182.255
INTERFACE_STATE[1]=up
DHCP_ENABLE[1]=0
#
ROUTE_DESTINATION[0]="default"
ROUTE_MASK[0]="25.255.255.0"
ROUTE_GATEWAY[0]="172.23.4.3"
ROUTE_COUNT[0]="1"
ROUTE_ARGS[0]=""
#
GATED=0
GATED_ARGS=""
RDPD=0
RARP=0
#

Here is my netstat -rn

Routing tables
Destination Gateway Flags Refs Interface Pmtu
127.0.0.1 127.0.0.1 UH 0 lo0 4136
172.23.4.144 172.23.4.144 UH 0 lan1 4136
172.23.182.21 172.23.182.21 UH 0 lan0 4136
172.23.4.0 172.23.4.144 U 2 lan1 1500
172.23.182.0 172.23.182.21 U 2 lan0 9000
default 172.23.4.3 UG 0 lan1 0


172.23.182.0 is the closed segment.

Any help appreciated.

[Rita C Workman]

Well...being tidy and proofing the first thing I see is that on your "default" ROUTE_MASK....do you really want 25.255.255.255.0 ?
Maybe 255.255.255.0 would work better..

Rgrds,
Rita

[Patrick Wallek]

I don't know if it'll make any difference, but your ROUTE_MASK[0]="25.255.255.0" is incorrect. It should be 255.255.255.0. You first octet is 25 instead of 255.

[Haim Lichaa]

That was a typo on my part. fixed it and didn't make any difference.

Any other suggestions?

[Ron Kinner]

If it works in one direction but not in the other then the problem is not a routing issue. We are talking tcp here so it has to be able to talk in both directions.

More like a permission/security problem or possibly a firewall filter or even a name lookup problem.

http://docs.hp.com/en/B2355-90147/ch02s13.html

Ron

By the way the Default route should have a mask of "" not 255.255.255.0 or just delete the whole line since it is not required.

[Haim Lichaa]

No that can't be the issue.
1. No firewall between the two.
2. One is a dd copy of the other.

[Jim Allan]

Does the traffic pass through a router inbetween?

Can you pass ICMP traffic between the two?

[Jim Allan]

Sorry I should've been more specific with the router question.

The problem could be asymmetric routing, in which the route back takes an alternative path (one doomed for failure).

Also try running a traceroute between them from both different sides and post results

[T G Manikandan]

you cannot assume a network issue,

why do not you have the latest patches for RCP.

[Haim Lichaa]

There isn't any router between the two.
I can ping from both to each other on both the closed and open Networks.
Traceroute looks as if it's trying to send a reply via the other interface, even though ping -o shows the contrary.

####
traceroute 172.23.182.21
traceroute: Warning: Multiple interfaces found; using 172.23.4.145 @ lan1
traceroute to 172.23.182.21 (172.23.182.21), 30 hops max, 40 byte packets
1 f18udb01 (172.23.4.144) 0.369 ms 0.103 ms 0.093 ms

#####
#ping -o 172.23.182.21
PING 172.23.182.21: 64 byte packets
64 bytes from 172.23.182.21: icmp_seq=0. time=0. ms

----172.23.182.21 PING Statistics----
1 packets transmitted, 1 packets received, 0% packet loss
round-trip (ms) min/avg/max = 0/0/0
1 packets sent via:
172.23.182.21 - f18udb01nb
172.23.182.22 - f18udb02nb


Network 172.23.182.0 is the closed network
Network 172.23.4.0 is the opened network

Thanks.

[Ron Kinner]

traceroute is broken. Can't be trusted unless you use the -i option.

Ping -o is better.

I noticed you used the IP address which has to work since the rcp tcp session was established.

Remember rcp uses TCP so there must be an exchange of SYN and Ack Packages before the session starts which means the routing has to be good. Even if it used UDP it would still have to send and receive packets in both directions.

Since there is no router and no firewall then we must either be looking at a name lookup problem or some sort of permission/security issue.

Are you sure that the second command is correct? I would think it would be either / or . but not both or perhaps you meant ./ instead. Do you have write permission in the root directory? Aren't you trying to name the file "."? Did you try giving it a new file name?
rcp remoteserver:/tmpfile junk
or perhaps
rcp remoteserver:/tmpfile localserver:/tmp/junk

What happens if you use the name in your ping -o? Also you need to do it from both boxes.

Ron

[Haim Lichaa]

I don't think it's a name resolution issue.
I tried the same with IPs as root.

#rcp 172.23.182.21:/etc/hosts /tmp
rcp: interrupted by signal 2
root@server1:[/root]
#ls -l /tmp/hosts
---------- 1 root sys 0 Jan 12 08:27 /tmp/hosts


The command just hung. I had to kill it.
As you can see an empty file is created.

I have the same problem with NFS mount over this dedicated (closed) LAN. I can mount ok, but when I start writing the mount hangs.

I've just opened a technical call to HP support.
Would be interesting to see how they handle it. If this thread continues and fails to arrive to a solution I will post the HP solution (if they arrive at one that is).

[doug mielke]

I'd try to ftp back as a test. RCP requires entries in .rhosts and/or /etc/hosts.equiv

if ftp works, then llok at the security re: to those 2 files.

[Haim Lichaa]

FTP connects ok, but when I try to get a file it hangs.

[Haim Lichaa]

PROBLEM RESOLVED!!!!

This was the lan0 MTU size.
9000 was causing massive fragmentation even though I'm configured for 1000Mb

Changing to 1500 resolved the issue.

Still don't know why the connection was only one way?

Thanks all for trying.